|Publication number||US7783629 B2|
|Application number||US 11/326,283|
|Publication date||Aug 24, 2010|
|Filing date||Jan 5, 2006|
|Priority date||Dec 13, 2005|
|Also published as||US7707204, US20070136280, US20070136281|
|Publication number||11326283, 326283, US 7783629 B2, US 7783629B2, US-B2-7783629, US7783629 B2, US7783629B2|
|Inventors||Hang Li, Jianfeng Gao, Yunbo Cao|
|Original Assignee||Microsoft Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (39), Non-Patent Citations (31), Referenced by (11), Classifications (8), Legal Events (2)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The present application is a divisional of and claims priority of U.S. patent application Ser. No. 11/302,560, filed Dec. 13, 2005, the content of which is hereby incorporated by reference in its entirety.
Some current studies have found that workers at many companies spend a significant amount of their time searching for information. A significant number of those workers also report that they find it very difficult to locate information that is required to do their jobs on their internal network systems (such as intranets). Assisting these types of people in accessing information (particularly at an enterprise level) is a challenging issue which many have attempted to address with conventional information retrieval techniques.
However, some conventional information retrieval techniques are not optimally suited to the job. For instance, workers often require not only general search capabilities, but also the capability to search for specific types of information, such as proper names, time schedules, etc.
One current approach to information retrieval is referred to as question answering. the methods used in implementing some current question answering systems are not normally developed to answer questions (or provide information) on a specific type of factoid, and thus, they are not tuned for the type. As a result, question answering systems do not perform well when searching for those types of factoids. Further, question answering systems are often relatively slow, because it is difficult to generate an index for them.
Another type of current information retrieval is known as document retrieval, which is conducted on the basis of a relevance measure that indicates the relevance of a set of documents to queries. Passages are sometimes retrieved, instead of entire documents, for ease of information access. The documents or passages often do not necessarily contain a specific type of information that was requested by the input query.
The discussion above is merely provided for general background information and is not intended to be used as an aid in determining the scope of the claimed subject matter.
A query and a factoid type selection are received from a user. An index of passages, indexed based on factoids, is accessed and passages that are related to the query, and that have the selected factoid type, are retrieved. The retrieved passages are ranked and provided to the user based on a calculated score, in rank order.
This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter. The claimed subject matter is not limited to implementations that solve any or all disadvantages noted in the background.
The present invention deals with searching information based on a selected factoid type. However, before describing the present invention in more detail, one illustrative environment in which the present invention can be used will be discussed.
Embodiments are operational with numerous other general purpose or special purpose computing system environments or configurations. Examples of well-known computing systems, environments, and/or configurations that may be suitable for use with various embodiments include, but are not limited to, personal computers, server computers, hand-held or laptop devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, telephony systems, distributed computing environments that include any of the above systems or devices, and the like.
Embodiments may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. Some embodiments are designed to be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules are located in both local and remote computer storage media including memory storage devices.
With reference to
Computer 110 typically includes a variety of computer readable media. Computer readable media can be any available media that can be accessed by computer 110 and includes both volatile and nonvolatile media, removable and non-removable media. By way of example, and not limitation, computer readable media may comprise computer storage media and communication media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by computer 110. Communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media. The term “modulated data signal” means a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. By way of example, and not limitation, communication media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared and other wireless media. Combinations of any of the above should also be included within the scope of computer readable media.
The system memory 130 includes computer storage media in the form of volatile and/or nonvolatile memory such as read only memory (ROM) 131 and random access memory (RAM) 132. A basic input/output system 133 (BIOS), containing the basic routines that help to transfer information between elements within computer 110, such as during start-up, is typically stored in ROM 131. RAM 132 typically contains data and/or program modules that are immediately accessible to and/or presently being operated on by processing unit 120. By way of example, and not limitation,
The computer 110 may also include other removable/non-removable volatile/nonvolatile computer storage media. By way of example only,
The drives and their associated computer storage media discussed above and illustrated in
A user may enter commands and information into the computer 110 through input devices such as a keyboard 162, a microphone 163, and a pointing device 161, such as a mouse, trackball or touch pad. Other input devices (not shown) may include a joystick, game pad, satellite dish, scanner, or the like. These and other input devices are often connected to the processing unit 120 through a user input interface 160 that is coupled to the system bus, but may be connected by other interface and bus structures, such as a parallel port, game port or a universal serial bus (USB). A monitor 191 or other type of display device is also connected to the system bus 121 via an interface, such as a video interface 190. In addition to the monitor, computers may also include other peripheral output devices such as speakers 197 and printer 196, which may be connected through an output peripheral interface 195.
The computer 110 is operated in a networked environment using logical connections to one or more remote computers, such as a remote computer 180. The remote computer 180 may be a personal computer, a hand-held device, a server, a router, a network PC, a peer device or other common network node, and typically includes many or all of the elements described above relative to the computer 110. The logical connections depicted in
When used in a LAN networking environment, the computer 110 is connected to the LAN 171 through a network interface or adapter 170. When used in a WAN networking environment, the computer 110 typically includes a modem 172 or other means for establishing communications over the WAN 173, such as the Internet. The modem 172, which may be internal or external, may be connected to the system bus 121 via the user input interface 160, or other appropriate mechanism. In a networked environment, program modules depicted relative to the computer 110, or portions thereof, may be stored in the remote memory storage device. By way of example, and not limitation,
By factoid type, the present discussion means individual types or categories of information, sometimes referred to as named entities, and including for example, place names, organization names, telephone numbers, times, dates, events, product names, city names, other proper names, etc. Operation of indexing system 202 is described in more detail below with respect to
Once the passages and scores are indexed by factoid type, they can be used to perform runtime searches with runtime system 204. Runtime system 204 includes passage retrieval component 250 and passage ranking component 252. Component 250 receives a query 256 and factoid type selection input 258 and retrieves related passages 251 from index 216. Component 252 scores and ranks passages 251 to provide ranked passages 282. This is described in greater detail below with respect to
Passage construction component 206 accesses documents 212 and constructs passages 207 from those documents 212 for each of the factoid types 214 to be indexed. Constructing the passages is indicated by block 302 in
Score computation component 208 computes a score 209 for each of the passages 207, based on how closely the passage is related to the various factoid types being indexed. Computing the score is indicated by block 304 in
Passage indexing and component 210 then indexes both the constructed passages 207, and the corresponding scores 209 which have been computed for each of those passages 207. Passage indexing component 210 places the passages and their scores 209 in passage index store 216. This is indicated by block 306 in
The indexed passages in store 216 are then used by system 204 to retrieve relevant passages based on an input query. However, before discussing the retrieval process in more detail, one embodiment of the system for constructing passages 207 from documents 212 will be described. This is shown at blocks 308 and 310 in
Passage construction component 206 first progressively selects each factoid type 214 for indexing. Component 206 then identifies all expressions of each factoid type in the corpus of documents 212 to be indexed. This is indicated by block 308 in
The ranking function φ(w,p) is now described in more detail. In one embodiment, for each content word w in the passage p, a vector of features is created, with m dimensions, and is denoted as (xw,p,1, . . . , xw,p,m), and a linear combination model is employed to compute φ(w,p) as follows:
where cj for j=1, . . . , m are weights of different features in the feature vector. The weights cj are trained as described below with respect to
Features of Word
Inverse document frequency (IDF) value of
Term frequency (TF) value of Word in original
TF * IDF value of Word;
Distribution of Word in original document,
measured in terms of entropy.
Part of speech of Word.
Features of passage
Type of Time in passage (e.g., “Aug. 1, 2000”,
“Aug., 2000”, or “2000”);
Bag of words in passage (i.e., the actual
words in the passage);
Size of passage.
Features on location
Minimum distance between Time and Word, in
terms of the number of words, sentences, and
paragraphs, respectively separating them;
Does Word occur in same sentence or paragraph
Does Word occur in the same partial sentence
(segmented by “,”) as Time?
Does Word occur one line below or above
current line only containing Time?
Does Word occur in metadata title?
Does Word occur in previous or next sentence
Does Word occur immediately before or after
Features on the relation of this expression
with other time expressions
Is there any other time expression in the same
sentence as Time?
Is there any other time expression closer to
the Word (in terms of word, sentence, and
It is assumed for the sake of
Component 206 then selects the current paragraph, which is the paragraph in document 212 that contains the selected factoid expression. This is indicated by block 342. Component 206 determines whether the length of the current passage is shorter than a threshold number, N, words. This is indicated by block 344. If the current paragraph has fewer words than the threshold number N, then component 206 chooses the entire current paragraph as the passage 207 for that factoid expression. This is indicated by block 346 in
Component 206 then determines whether the current paragraph contains only the factoid expression. In other words, if the factoid expression is a time of day (say 8:00 pm), and that time of day resides as its own paragraph, then component 206 adds the immediately preceding and following paragraphs to the passage 207 for that expression. Determining whether the paragraph contains only the factoid expression, and, if so, adding the preceding and following paragraphs to the passages are indicated by blocks 348 and 350 in
If, at block 348, it is determined that the current paragraph does not contain only the factoid expression, then processing continues at block 354. This is discussed below.
If, at block 344, it is determined that the length of the current paragraph is not shorter than the threshold number N words, then component 206 adds the current sentence (the sentence that contains the factoid expression) to the passage 207. This is indicated by block 352 in
Once the determination has been made to add the sentence containing, or paragraphs surrounding, the factoid expression, processing continues at block 354 which is optional. In the embodiment in which block 354 is used, component 206 adds words from the title of the document containing the factoid expression to the passage 207 corresponding to the factoid expression. In one embodiment, the documents 212 can be HTML documents in which the title is identified in metadata. In that case, the words from the metadata identifying the title in the HTML document 212 can be added. In other embodiments, the text from the title can be added to the passage 207.
Block 356 is also optional. In the embodiment in which block 356 is used, component 206 determines whether the factoid expression is in a table. If so, then component 206 adds the words from the caption of the table to the passage 207.
Once these items have been added to passage 207, component 206 determines whether the passage has reached the threshold N words or more. This is indicated by block 358 in
If, at block 358, it is determined by component 206 that the passage has not reached the threshold number N words, then component 206 determines whether, in the processing at block 344, a sentence was added, or paragraphs where added, to the passage. This indicated by block 362. If a sentence was added (as at block 352) then the preceding and following sentences are added to the passage 207. This is indicated by block 364. Processing again reverts back to block 358 to determine whether the passage 207 contains at least N words. If not, processing again follows through blocks 362 and 364 in which more preceding and following sentences (preceding and following the current passage) are added to the passage. This continues until enough preceding and following sentences are added to the passage such that the passage includes at least N words.
If, at block 362, it is determined that in the above processing in
After receiving the query 256 and factoid type selection 258, passage retrieval component 250 accesses passage-index store 216 to obtain access to all passages of the selected factoid type 258. Recall that the passages are indexed by factoid type so they can be easily accessed given the factoid type selection 258. This is indicated by block 272 in
Passage retrieval component 250 then retrieves passages 251 from index 216 that are related to the search query 256 and have the selected factoid type 258. This is indicated by block 274 in
In scoring and ranking the retrieved passages 251, a number of different scoring systems can be used. In one illustrative embodiment, given a search query q, and each passage 251 retrieved with q, a ranking function s(q,p) is defined as follows:
where w stands for a keyword (content word) appearing in both the query q and the passage p. The function in Eq. 1 represents the likelihood that the factoid expression of the passage p is the correct answer of an event described by the query q. For instance, where the factoid expression is a time expression, the function represents the likelihood that the time expression of the passage p is the correct answer of an event described by the query q.
In order to rank the passages, the score from Eq. 1 is calculated for each passage 251, while looking up from the index 216 the pre-computed scores of all related Φ(w,p) pairs. These are the scores which were pre-calculated offline as described above. Then, the passages are sorted in descending order of the score computed in Eq. 1. Outputting the ranked passages 282 is indicated by block 280.
The ranked passages can then be used by a user, as desired. This is indicated by block 284 in
Then, the instances are used to construct a model for ranking the passages. The model utilizes information on the position relation between the factoid expression and keywords in the query, the characteristics of the keyword, the characteristics of the passages, etc. This is indicated by block 410.
(p1,y1), . . . , (pi,yi), . . . , (p1,y1) Eq. 3
where pi denotes a passage and yi denotes a label having the value 1 or 0 (denoting a positive or negative instance). For each keyword (content word) wk appearing in both the query q and the passage pi, the corresponding feature vector (xw,p,1, . . . , xw,p,m) is created. Next, the value Φ(wkpi) (as set out in Eq. 2) is calculated, as is the score s(q,pi) as set out in Eq. 1.
The passages p1, . . . , pl are then ranked for each query q based on their scores calculated as described in Eq. 1. Since there are answers (labels) y1, . . . , yl for each query q, the total ranking results can be evaluated. Training the model (or machine learning) is a process of finding the optimal weight vector (c1, . . . , cm) for achieving the optimal ranking results in terms of an evaluation measure.
A plurality of different weight values are then tried for cj, and ranking is performed using the training data. The performance of the ranking is then evaluated as indicated by block 556. In one illustrative embodiment, the performance of the ranking is evaluated in terms of Mean Average Precision (MAP). Mean Average Precision is defined as a Mean of the Average Precision over queries and the Average Precision for each query is defined as follows:
Where p(j) denotes precision over the top j results, and Δr(j) is a change in recall from j−1 to j.
The best performance and corresponding weight value cj is recorded as indicated by block 558 and cj is reset to its original value t. This is indicated by block 560.
If the best performance in terms of Mean Average Precision for the value cj is an improvement by more than a threshold amount over the previous value of cj, then the new value for cj is adopted (that value which corresponds to the best performance just computed). This is indicated by blocks 562 and 564 in
Processing then progresses to block 566 where it is determined whether there are more dimensions to check (i.e., whether j<m). If not, the weight values for all of the dimensions of the feature vector have been computed and the process concludes. However, if there are more values of cj to be computed, then the weight value for the next dimension in the feature vector (i.e., j=j+1 for cj) is selected as illustrated at block 568 and the process continues at block 554.
In accordance with one embodiment, the particular measure used in passage ranking is a distance measure. For instance, if the key words in the query match the passage, and the matched keywords in the passage are close to the factoid expression in the passage, then it is likely that the factoid expression is the answer to the query. One such method employs the following distance function as an exemplary ranking function:
where w denotes a keyword, t denotes time, d is the minimum distance between t and w, and a is a parameter that can be empirically set.
In accordance with yet another embodiment, retrieved passages 251 are ranked by using a voting algorithm. In other words, each passage 251 has a score and each passage 251 may support more than one factoid expression. In one embodiment, voting logic is used to weight the scores to determine which time expression is most strongly supported by the passage 251. For instance, assume a first retrieved passage 251 supports a first expression with a weight of 0.5 and a second passage 251 supports the first expression with a weight of 0.2, but a third passage 251 also supports a second expression with a weight of 0.6. Then, the first expression will be weighted higher than the second expression, because it has a total weight of 0.7, while the second expression has a total weight of 0.6. This is true even though the score for the second expression is higher than the highest score for the first expression. The sum of scores for the second expression is still lower than the sum of all the scores for the first expression. In accordance with another embodiment, voting is simply conducted using the numbers of occurrences of the factoid expression in the passage.
In either voting method, the particular factoid expressions whose ratios of keywords matched to passages is less than a predetermined value β, or which are not ranked in the top k passages, are discarded. In one embodiment, β is approximately 0.6 while k is approximately 13. However, any of a wide variety of other values of β and k can be used and the values can be empirically determined. It has been observed that using one or both of these voting techniques after ranking can further enhance the performance of passage ranking. Similarly, the use of one or more of these voting algorithms can improve both the rule-based passage construction method and the passage ranking methods described above with respect to
Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5696962 *||May 8, 1996||Dec 9, 1997||Xerox Corporation||Method for computerized information retrieval using shallow linguistic analysis|
|US5983218 *||Jun 30, 1997||Nov 9, 1999||Xerox Corporation||Multimedia database for use over networks|
|US6026388 *||Aug 14, 1996||Feb 15, 2000||Textwise, Llc||User interface and other enhancements for natural language information retrieval system and method|
|US6078916||Mar 12, 1998||Jun 20, 2000||Culliss; Gary||Method for organizing information|
|US6269368 *||Oct 16, 1998||Jul 31, 2001||Textwise Llc||Information retrieval using dynamic evidence combination|
|US6446061||Jun 30, 1999||Sep 3, 2002||International Business Machines Corporation||Taxonomy generation for document collections|
|US6678694 *||Nov 8, 2000||Jan 13, 2004||Frank Meik||Indexed, extensible, interactive document retrieval system|
|US6772149 *||Sep 23, 1999||Aug 3, 2004||Lexis-Nexis Group||System and method for identifying facts and legal discussion in court case law documents|
|US6948123||Mar 20, 2002||Sep 20, 2005||Fujitsu Limited||Multimedia information arranging apparatus and arranging method|
|US7089226 *||Jun 28, 2001||Aug 8, 2006||Microsoft Corporation||System, representation, and method providing multilevel information retrieval with clarification dialog|
|US7181438 *||May 30, 2000||Feb 20, 2007||Alberti Anemometer, Llc||Database access system|
|US7246060||Nov 6, 2001||Jul 17, 2007||Microsoft Corporation||Natural input recognition system and method using a contextual mapping engine and adaptive user bias|
|US7249013 *||Mar 11, 2003||Jul 24, 2007||University Of Southern California||Named entity translation|
|US7251648 *||Jun 28, 2002||Jul 31, 2007||Microsoft Corporation||Automatically ranking answers to database queries|
|US7269545 *||Mar 30, 2001||Sep 11, 2007||Nec Laboratories America, Inc.||Method for retrieving answers from an information retrieval system|
|US7529731 *||Jun 29, 2004||May 5, 2009||Xerox Corporation||Automatic discovery of classification related to a category using an indexed document collection|
|US7558792 *||Jun 29, 2004||Jul 7, 2009||Palo Alto Research Center Incorporated||Automatic extraction of human-readable lists from structured documents|
|US7707204 *||Dec 13, 2005||Apr 27, 2010||Microsoft Corporation||Factoid-based searching|
|US20020048269||Aug 3, 2001||Apr 25, 2002||Hong Jack L.||Intelligent demand driven recognition of URL objects in connection oriented transactions|
|US20030088410 *||Nov 6, 2001||May 8, 2003||Geidl Erik M||Natural input recognition system and method using a contextual mapping engine and adaptive user bias|
|US20030208477 *||May 2, 2002||Nov 6, 2003||Smirniotopoulos James G.||Medical multimedia database system|
|US20030229604 *||Jun 10, 2002||Dec 11, 2003||Microsoft Corporation||Large margin perceptrons for document categorization|
|US20050108266 *||Nov 14, 2003||May 19, 2005||Microsoft Corporation||Method and apparatus for browsing document content|
|US20050114324 *||Sep 14, 2004||May 26, 2005||Yaron Mayer||System and method for improved searching on the internet or similar networks and especially improved MetaNews and/or improved automatically generated newspapers|
|US20050216426 *||May 20, 2002||Sep 29, 2005||Weston Jason Aaron E||Methods for feature selection in a learning machine|
|US20060136411 *||Dec 21, 2004||Jun 22, 2006||Microsoft Corporation||Ranking search results using feature extraction|
|US20060174211 *||Mar 31, 2006||Aug 3, 2006||Microsoft Corporation||Methods, apparatus and data structures for providing a user interface which facilitates decision making|
|US20060212142 *||Mar 15, 2006||Sep 21, 2006||Omid Madani||System and method for providing interactive feature selection for training a document classification system|
|US20060224569||Jul 11, 2005||Oct 5, 2006||Desanto John A||Natural language based search engine and methods of use therefor|
|US20060248049 *||Apr 27, 2005||Nov 2, 2006||Microsoft Corporation||Ranking and accessing definitions of terms|
|US20070094285 *||Oct 21, 2005||Apr 26, 2007||Microsoft Corporation||Question answering over structured content on the web|
|US20070112720 *||Nov 14, 2005||May 17, 2007||Microsoft Corporation||Two stage search|
|US20070136246 *||Dec 28, 2005||Jun 14, 2007||At&T Corp.||Answer determination for natural language questioning|
|US20070136280 *||Dec 13, 2005||Jun 14, 2007||Microsoft Corporation||Factoid-based searching|
|US20070140472 *||Sep 28, 2006||Jun 21, 2007||Hon Hai Precision Industry Co., Ltd.||xDSL MODEM AND METHOD FOR STABILIZING CONNECTION OF xDSL MODEM|
|US20070179776||Jan 27, 2006||Aug 2, 2007||Xerox Corporation||Linguistic user interface|
|US20070260580 *||Nov 17, 2006||Nov 8, 2007||Nosa Omoigui||Information nervous system|
|US20080263003 *||Dec 22, 2005||Oct 23, 2008||Tomasz Imielinski||Method and apparatus for increasing query traffic to a web site|
|US20100082639 *||Sep 30, 2008||Apr 1, 2010||Microsoft Corporation||Processing maximum likelihood for listwise rankings|
|1||Abraham Ittycheriah, Salim Roukos, IBM's Statistical Question Answering System-TREC II. TREC 2002.|
|2||Charles L. A. Clarke, Gordon V. Cormack, Thomas R. Lynam, C. M. Li, G. L. McLearn, Web Reinforced Question Answering (MultiText Experiments for TREC 2001). TREC 2001.|
|3||Cody C. T. Kwok, Oren Etzioni, Daniel S. Weld, Scaling question answering to the Web. WWW-2001: 150-161.|
|4||Dragomir R. Radev, Weiguo Fan, Hong Qi, Harris Wu, Amardeep Grewal, Probabilistic question answering on the web. WWW 2002: 408-419.|
|5||Eduard H. Hovy, Laurie Gerber, Ulf Hermjakob, Michael Junk, Chin-Yew Lin, Question Answering in Webclopedia. TREC 2000.|
|6||Eric Brill, Susan Dumais and Michele Banko, An Analysis of the AskMSR Question-Answering System, EMNLP 2002.|
|7||Gerard Salton, James Allan, Chris Buckley, Approaches to Passage Retrieval in Full Text Information Systems. SIGIR 1993: 49-58.|
|8||Gordon V. Cormack, Charles L. A. Clarke, D. I. E. Kisman, Christopher R. Palmer, Fast Automatic Passage Ranking (MultiText Experiments for TREC-8). TREC 1999.|
|9||Hui Yang, Hang Cui, Mstislav Maslennikov, Long Qiu, Min-Yen Kan, Tat-Seng Chua, Qualifier in TREC-12 QA Main Task. TREC 2003: 480-488.|
|10||Jay M. Ponte, W. Bruce Croft, A Language Modeling Approach to Information Retrieval. SIGIR1998: 275-281.|
|11||John M. Prager, Eric W. Brown, Anni Coden, Dragomir R. Radev, 2000. Question-answering by predictive annotation. SIGIR 2000: 184-191.|
|12||Li Zhang, Yue Pan, Tong Zhang, Focused named entity recognition using machine learning. SIGIR 2004: 281-288.|
|13||Lucian Vlad Lita, Monica Rogati, Jaime G. Carbonell, Cross Lingual QA: A Modular Baseline in CLEF 2003. CLEF 2003: 535-540.|
|14||Pending U.S. Appl. No. 11/302,560, filed Dec. 13, 2005, entitled Factoid-Based Searching.|
|15||Stephen E. Robertson, Steve Walker, Micheline Hancock-Beaulieu, Mike Gatford, A. Payne, Okapi at TREC-4. TREC 1995.|
|16||Susan Feldman, The high cost of not finding information, KMWorld-vol. 13, Issue 3.|
|17||Thomas S. Morton, Using Coreference in Question Answering, TREC1999, 1999.|
|18||Typed Enterprise Search: Investigation on Time Search, http://msranl:81/Information%20Extraction/Reports/Search%20for%20Time-draft27-.pdf , Submission to SIGIR 2005.|
|19||Typed Enterprise Search: Investigation on Time Search, http://msranl:81/Information%20Extraction/Reports/Search%20for%20Time—draft27—.pdf , Submission to SIGIR 2005.|
|20||U.S. Appl. No. 11/302,560 Advisory Action mailed Apr. 1, 2009.|
|21||U.S. Appl. No. 11/302,560 Amendment filed Aug. 1, 2008.|
|22||U.S. Appl. No. 11/302,560 Amendment filed Mar. 24, 2009.|
|23||U.S. Appl. No. 11/302,560 Amendment filed Oct. 27, 2009.|
|24||U.S. Appl. No. 11/302,560 Notice of Allowance mailed Jan. 25, 2010.|
|25||U.S. Appl. No. 11/302,560 Office Action mailed Dec. 19, 2007.|
|26||U.S. Appl. No. 11/302,560 Office Action mailed Feb. 25, 2008.|
|27||U.S. Appl. No. 11/302,560 Office Action mailed Jul. 31, 2009.|
|28||U.S. Appl. No. 11/302,560 Office Action mailed Nov. 18, 2008.|
|29||U.S. Appl. No. 11/302,560 Response to Restriction Requirement filed Jan. 17, 2008.|
|30||William Morgan, Warren Greiff, John Henderson, Direct Maximization of Average Precision by Hill-Climbing, with a Comparison to a Maximum Entropy Approach, HLT-NAACL 2004, pp. 93-96.|
|31||Xiaoyong Liu, W. Bruce Croft, Passage retrieval based on language models. CIKM 2002: 375-382.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US9183430 *||Oct 3, 2013||Nov 10, 2015||Quanta Computer Inc.||Portable electronic apparatus and interactive human face login method|
|US9367490||Jun 13, 2014||Jun 14, 2016||Microsoft Technology Licensing, Llc||Reversible connector for accessory devices|
|US9384334||May 12, 2014||Jul 5, 2016||Microsoft Technology Licensing, Llc||Content discovery in managed wireless distribution networks|
|US9384335||May 12, 2014||Jul 5, 2016||Microsoft Technology Licensing, Llc||Content delivery prioritization in managed wireless distribution networks|
|US9430667||May 12, 2014||Aug 30, 2016||Microsoft Technology Licensing, Llc||Managed wireless distribution network|
|US9477625||Mar 31, 2016||Oct 25, 2016||Microsoft Technology Licensing, Llc||Reversible connector for accessory devices|
|US9614724||Apr 21, 2014||Apr 4, 2017||Microsoft Technology Licensing, Llc||Session-based device configuration|
|US9639818 *||Aug 30, 2013||May 2, 2017||Sap Se||Creation of event types for news mining for enterprise resource planning|
|US20120011112 *||Jul 6, 2010||Jan 12, 2012||Yahoo! Inc.||Ranking specialization for a search|
|US20150023569 *||Oct 3, 2013||Jan 22, 2015||Quanta Computer Inc.||Portable electronic apparatus and interactive human face login method|
|US20150066552 *||Aug 30, 2013||Mar 5, 2015||Sap Ag||Creation of event types for news mining for enterprise resource planning|
|U.S. Classification||707/723, 707/731, 707/748, 707/728, 707/749|
|Jan 28, 2014||FPAY||Fee payment|
Year of fee payment: 4
|Dec 9, 2014||AS||Assignment|
Owner name: MICROSOFT TECHNOLOGY LICENSING, LLC, WASHINGTON
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:MICROSOFT CORPORATION;REEL/FRAME:034543/0001
Effective date: 20141014